tomaarsen
@tomaarsen
Followers
2,349
Following
203
Media
112
Statuses
505
Sentence Transformers, SetFit & NLTK maintainer Machine Learning Engineer at 🤗 Hugging Face
Netherlands
Joined December 2023
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Rogan
• 779455 Tweets
Beyoncé
• 643212 Tweets
Hybe
• 547556 Tweets
Big Winner
• 466034 Tweets
最後の最後
• 461622 Tweets
CWR On ONE BKK Stage
• 237259 Tweets
#PBBGen11BigNight
• 222844 Tweets
Fyang
• 213821 Tweets
SB19 ANIMBERSARY CONCERT
• 205777 Tweets
Halloween with DAOU
• 131235 Tweets
CONSTRUCTIVE DREAMSCAPE 1
• 120464 Tweets
FY RECAP BLANK SS2EP5
• 109703 Tweets
AioonMay Be Together
• 96570 Tweets
国民民主
• 60741 Tweets
#祝NIKKE2周年生放送
• 57397 Tweets
#SixTONESANN
• 47782 Tweets
TXT SANCTUARY PREVIEW
• 46810 Tweets
#UFC308
• 43925 Tweets
シンデレラ
• 37059 Tweets
石丸さん
• 31672 Tweets
石丸伸二
• 31447 Tweets
玉木さん
• 29651 Tweets
Haaland
• 28199 Tweets
ホークス
• 24928 Tweets
早じまい
• 20428 Tweets
経費節減
• 19314 Tweets
Southampton
• 18133 Tweets
Leal
• 13391 Tweets
Pinned Tweet
📣 Sentence Transformers v3.2.0 is out, marking the biggest release for inference in 2 years! 2 new backends for embedding models: ONNX (+ optimization & quantization) and OpenVINO, allowing for speedups up to 2x-3x + Static Embeddings for 500x speedups at 10-20% accuracy cost
🧵
9
80
434
‼️Sentence Transformers v3.0 is out! You can now train embedding models with multi-GPU training, bf16 support, loss logging, callbacks & much more. I also release 50+ datasets to train on & much more.
Learn how to use the new Trainer here:
Details in 🧵
9
103
425
Embedding Quantization is here! 25x speedup in retrieval; 32x reduction in memory usage; 4x reduction in disk space; 99.3% preservation of performance🤯 The sky is the limit. Read about it here: More info in 🧵
Binary and Scalar Embedding Quantization for Significantly Faster & Cheaper Retrieval
huggingface.co
5
93
375
🔥Sentence Transformers v2.4.0 is released! It introduces Matryoshka Embedding models (training & inference), 2 new state-of-the-art loss functions, prompt templates, instructor model support & more. See the🧵
6
93
375
Model2Vec distills a fast model from a Sentence Transformer by passing its vocabulary through the model, reducing embedding dims via PCA and applying Zipf weighting.
Inference with the resulting static embeddings are lightning-fast, e.g. 10k texts/sec:
🧵
4
50
241
Label a few examples; train a small classifier; outperform LLMs 500x larger & slower for your classification task. Sounds difficult? It's not.
Check out this blogpost using Argilla, distilabel, and SetFit to try it yourself:
2
31
231
🏎️ I've just finished uploading ~1000 ONNX models & ~100 OpenVINO models for a ton of Sentence Transformers models.
Use them with: `SentenceTransformer("all-MiniLM-L6-v2", backend="onnx", model_kwargs={"file_name": "model_O4.onnx"})` for 1.3x to 2.5x speedups!
🧵
5
24
199
Sentence Transformers v3.1 is out! Featuring a hard negatives mining utility to get better models out of your data, a new strong loss function, training with streaming datasets, custom modules, bug fixes, small additions and docs changes.
Release notes:
🧵
5
34
175
🎉SetFit v1.1.0 is out! Training efficient classifiers on CPU or GPU now uses the Sentence Transformers Trainer, and we resolved a lot of issues caused by updates of third-party libraries (like Transformers).
Full release notes:
More details in 🧵
2
28
169
ONNX, Quantization, OpenVINO, Optimization, Lower precision, etc. are all rather confusing. Luckily, for Sentence Transformers you can just follow this flowchart based on my personal benchmarks.
Docs:
6
25
169
78.17% -> 93.40% accuracy by finetuning an embedding dataset on a small synthetic dataset with Sentence Transformers v3! Great work 👏
Thanks to
@tomaarsen
's work in the latest Sentence Transformers release, training custom models is easier than ever.
With improved training support and synthetic data for fine-tuning, you can build a model in less than a day.
Example here👇:
1
8
77
0
11
144
Biggest release of the week: just released 'mxbai-embed-large-v1', a top embedding model outperforming all equivalently sized models such as 'bge-large-en-v1.5'. It's Apache-2.0 licensed, i.e. commercially viable.
Model link:
🧵1/5
3
22
140
Jack's excellent, state-of-the-art CDE-small-v1 is now compatible with Sentence Transformers with a really simple interface, check it out:
man, the huggingface team is *cracked*
two days after i released my contextual embedding model, which has a pretty different API,
@tomaarsen
implemented CDE in sentence transformers
you can already use it
implementation was not at all trivial; those people just work fast
18
101
1K
2
10
138
As far as attention-free, static embedding models solutions go, Model2Vec is the new absolute king, beating out BPEmb and GLoVe on various domains.
Learn how it works here:
P.s.: Integrated in Sentence Transformers, LangChain, LlamaIndex, Haystack, etc.
2
23
197
Thrilled to share that I've joined
@huggingface
🤗 as a Machine Learning Engineer tasked with maintaining the awesome Sentence Transformers project!
It's high time to bring modern training functionality to finetuning embedding models!⚡️
5
11
124
GLiNER has new Apache 2.0 models for efficient, cheap and high quality information extraction.
> 2 English models: small (152M) and medium (195M)
> 1 Multilingual model (288M)
> 25x-50x smaller and faster than any 7b LLM
> 3 lines of code
Demo on CPU:
3
33
122
The long-awaited Sentence Transformers v2.3.0 is now released! It contains a ton of bug fixes, performance improvements, loading custom models, more efficient loading, a new strong loss function & more!
Check out the release notes:
Or this 🧵below:
5
24
107
Alongside our blogpost on Embedding Quantization, we released a useful demo showcasing that it allows for <0.1s retrieval across all of Wikipedia (41 million texts) while using 32x less memory than normal retrieval. E.g. <10GB memory rather than 160GB 💸
3
9
96
Big update for the Massive Text Embedding Benchmark (MTEB) intended to simplify finding a good embedding model! Model filtering, search, memory usage, model size in parameters.
The updated leaderboard:
Details in 🧵:
2
22
91
Synthetic data used to improve an embedding model from 78.1% accuracy -> 93.4% accuracy with Sentence Transformers finetuning! Learn how to do this yourself ⏬ Great work, Daniel!
Do you need a dataset to train a custom sentence transformer model? I've created a pipeline for using an LLM to create a synthetic dataset you can directly use for fine-tuning/training a Setence Transformers model.
*Link in next tweet
3
19
103
5
9
84
I've just shipped the Sentence Transformers v3.1.1 patch release, fixing the hard negatives mining utility for some models. This utility is extremely useful to get more performance out of your embedding training data.
Release notes:
More info in 🧵
4
13
83
Sentence Transformers just crossed ⭐15k stars on GitHub!
Great timing too, we're about to ship a huge update: 2 new backends for computing embeddings via ONNX and OpenVINO, natively in Sentence Transformers. Expect faster inference very soon 🏎️!
1
3
82
released 3 state-of-the-art open text reranker models: fully Apache 2.0 & outperforming current top models such as bge-reranker-large and cohere-embed-v3 on BEIR datasets. All models are ready to use on the
@huggingface
Hub!
More details in 🧵
3
18
65
Jina AI has just released jina-embeddings-v3, a multilingual text embedding model optimized for a wide range of NLP tasks! It ranks very strongly on MTEB (
#2
across all tasks for <1B param models).
Model link🤗:
Details in 🧵
1
7
62
2 new tiny Apache 2.0 reranker models just got released by
@JinaAI_
. Despite their small size/latency, they perform competitively on benchmarks, reportedly outperforming bge-reranker-base and mxbai-rerank-base on MTEB Retrieval.
Models:
Details in 🧵
3
15
61
Snowflake has just shook up the MTEB Retrieval leaderboard with 5 new model releases:
> 23, 33, 110, 137 & 335M parameters
> Apache 2.0 license
> SOTA performance for their sizes/speeds
> 512 & 8192 seq length
Technical report coming soon.
Model link:
3
8
60
We just crossed 9000 public Sentence Transformer models on the
@huggingface
Hub! Big ups to everyone finetuning their own models 🤗
3
6
53
There's new embedding models for identifying instances where LLM instructions and responses don't align, by
@mrm8488
. They're useful for quality assurance, LLM training dataset filtering, retrieval, reward model in RLHF, etc.
@huggingface
:
Details in 🧵
2
5
53
Sentence Transformers just reached 14k stars! Just in time for the upcoming v3.0 update 👀It'll be the biggest update since the inception of the project. More details coming soon!
0
0
43
Sentence Transformers v2.7.0 is out! Featuring a new loss function, easier Matryoshka model inference & evaluation, CrossEncoder improvements & Intel Gaudi2 Accelerator support.
Release notes:
Or read the details in 🧵
1
8
37
Just published Sentence Transformers v3.0.1: the first patch release since v3 from last week. It introduces gradient checkpointing, pushing model checkpoints to Hugging Face while training, model card improvements and fixes.
Release notes:
Details in 🧵
1
2
37
📈 There's almost 4000 open source Sentence Transformers models on the Hugging Face Hub right now! Open source for the win❤️
5
8
34
Recently,
@mixedbreadai
and
@deepset_ai
collaborated on a SOTA German text embedding model, outperforming multilingual-e5-large and jina-embeddings-v2-base-de.
Link:
Details:
- 478M parameters: small enough to run on CPU and GPU
🧵
1
7
34
The last Sentence Transformers release introduced GISTEmbedLoss by
@avsolatorio
, which allows for training models that outperform those trained by the wildly popular in-batch negatives loss (MultipleNegativesRankingLoss). Learn about it in this 🧵(links at the bottom):
3
8
33
Absolutely loving the flexibility of Sentence Transformers v3 for training embedding models - allows for much easier paper reproductions.
2
2
29
@jxmnop
I'm glad I could showcase the flexibility of the Custom Modules from the latest Sentence Transformers v3.1 update here!
The final API is really simple; I'm quite happy with it.
0
0
29
🤗 The long-awaited full release of SetFit is finally out! SetFit v1.0.0 brings an all-new Trainer, TrainingArguments, logging, evaluation, integrations, callbacks, model cards, docs & more! 1/6
1
9
27
Nice! You can now link from older
@huggingface
models to their newer counterparts with a neat widget. As simple as adding the "new_version" metadata in the README.
1
4
25
Carbon emissions are automatically included with Sentence Transformer training if you install codecarbon: pip install codecarbon
E.g. see
New feature on the Hub!
☁️ Carbon emissions emitted during training now show up on the model card!
(requires model authors to fill that info first)
Hopes it will prompt more people to show the carbon emissions of their model training! 🌍
Thanks a lot to the team who pushed
1
8
29
0
7
25
Matryoshka embedding models can produce useful embeddings of various dimensions, which can heavily speed up downstream tasks like retrieval (e.g. for RAG).
Check out our blogpost with all of the details:
1
3
20
You can now finetune embeddings models with Sentence Transformers & AutoTrain without writing any code! Works locally, in Google Colab, on any cloud or via Hugging Face Spaces. Check it out!
🚨 NEW TASK ALERT 🚨
AutoTrain now supports fine-tuning of sentence transformer models 💥 Now, you can improve and customize your RAG or retrieval models without writing a single line of code 🤗
✅ Supports multiple types of sentence transformers training and finetuning
✅ CSV /
3
13
51
1
4
19
The 2000th public SetFit model was uploaded this morning! Awesome to see so many people enjoy SetFit for training efficient classifiers with very little training data 🤗
1
3
19
@bo_wangbo
@JinaAI_
Also, I would like to implement ColBERT training into Sentence Transformers (based on the HF Trainer with MultiGPU, bf16, callbacks + integrations, useful model cards, etc.), so I'm looking forward to your findings & promising loss functions there.
1
2
19
I'm currently in the process of uploading over 1000 ONNX (normal, optimized, quantized) and OpenVINO models, across all models under
🧵
1
0
17
Just crossed 2000 followers on here - many thanks!
I'll be sure to keep you all up to date on the latest in the world of Sentence Transformers and SetFit, as well as inform on embedding models and information retrieval/search systems in general!
Looking forward to 3k 🤗
0
0
16
We also reduced the dependencies & made many more small changes. See the release notes for all of this information in more detail:
I'm looking forward to seeing your models pop up on the Hub!🤗See you in the next release!
1
0
16
@huggingface
@mixedbreadai
This quantization is included in today's v2.6.0 Sentence Transformers release. Read the release notes here:
1
1
16
SetFit is extremely well suited for zero-shot text classification, and often outperforms much larger (and slower) zero-shot models on the Hugging Face Hub!
Check out our new how-to guide for zero-shot text classification here:
1
2
15
I really think that finetuned Static Embeddings + a cross-encoder reranker can be a very solid solution for efficient search. More info on that soon!
See here a finetuned Static Embedding model as an example:
🧵
1
1
15
CoSENTLoss is a new drop-in replacement of the popular CosineSimilarityLoss, while produces a stronger training signal. AnglELoss is another variant which uses a different similarity function to avoid vanishing gradients. See the loss docs for more info:
1
0
15
Read more about how to speed up Sentence Transformer models with lower precision, ONNX, OpenVINO, etc. in the "Speeding up Inference" documentation:
1
2
15
@mixedbreadai
released a🪆2D Matryoshka text embedding model. Such models have two notable properties: Adaptable embedding size & adaptable layer counts. It allows you to speed up both inference & all post-processing (e.g. retrieval).
Model link:
See 🧵
1
4
14
P.s. here's a gist for how to export all of these models:
0
2
14
An excellent training script for embedding models via the new Sentence Transformers v3 by
@mrm8488
. Try it out yourself!
🚀 Just out: Sentence-Transformers 3 is transforming the game! Kudos to
@tomaarsen
for the stellar update. 🌟
🔥 NEW FEATURE: Train your own Matryoshka embedding models! Want to dive in? I've set up a Colab notebook to get you started right away. Check it out and start creating
0
12
83
1
3
13
1️⃣ ONNX Backend: This backend uses the ONNX Runtime to accelerate model inference on both CPU and GPU, reaching up to 1.4x-3x speedup depending on the precision. We also introduce 2 helper methods for optimizing and quantizing models for (much) faster inference.
1
0
13
Read the blogpost below for a wonderful beginner-friendly introduction to Hugging Face Spaces to make your own AI/ML Web Apps!
getting to deploy open source, machine learning demos on
@huggingface
spaces really is such a perk of my job
if you want to learn how you can do this, i created a blog post for beginners :)
ps it uses
@Gradio
5
20
92
2
3
13
I integrated it in Sentence Transformers, so you can save and load all Model2Vec models with Sentence Transformers directly, also via LangChain, LlamaIndex, Haystack, etc.
ST Docs:
ST Release notes:
🧵
1
2
12
The Massive Text Embedding Benchmark (MTEB) is being extended to become massively multilingual. Everyone is invited to contribute & co-author an upcoming publication.
📜 Details:
0
3
12
Since yesterday's Sentence Transformers v3.0 update, distributed training of embedding models (for RAG, retrieval, semantic similarity, etc.) is now a breeze. You can expect some serious speedups when scaling the number of GPUs. Usage in 🧵
1
1
12
.
@huggingface
and
@mixedbreadai
announce embedding quantization: a post-processing technique for embeddings that results in massive cuts in costs for retrieval. E.g., rather than needing 200GB, we can search Wikipedia in this demo with just 5GB of RAM: 🧵
2
2
12
The new SetFit v1.0.0 release also brings SetFitABSA: Few-Shot Aspect Based Sentiment Analysis! ABSA is like Sentiment Analysis, except it tells you which parts people were happy/unhappy about, it's extremely useful!
Check out the blogpost:
1/3
1
3
12
Check out the full release notes for much more details:
Or read about Speeding up Inference in the new documentation page:
Looking forward to some solid feedback on this release!
0
4
11
AutoTrain now supports finetuning embedding models using Sentence Transformers! With other words: embedding models for your data without having to write any code.
Details:
More in 🧵
1
1
10
Recently, consensus has developed that larger sequence lengths result in notably worse embeddings, so this model uses a more reasonable 512.
🧵3/5
1
1
10
@1littlecoder
Embedding models are explicitly trained such that cosine similarity becomes a strong measure of semantic similarity. For all real-world embedding models, the findings of this paper do not apply at all. You can keep safely using cosine similarity.
1
0
11
@n0riskn0r3ward
@nvidia
To not get you too excited, there are a few concerns with this model at this point. The 1st place is mostly due to its high score on classification (87.35 vs 81.49 for
#2
), which is because it scores unexpectedly high on a few of the datasets, notably the EmotionClassification.
2
0
11
@eugeneyan
It seems rather challenging to access the underlying ClueWeb22 dataset (), but I would love love love to get this dataset on the Hub and in here:
1
1
10
🔥By applying optimum-intel we can get a 3.5x increase in throughput for SetFit text classification models on CPUs.
It applies quantization using the Intel Neural Compressor (INC), resulting in higher throughput on CPUs than with torch on GPUs.
Notebook:
2
2
10
Usage is as simple as `SentenceTransformer("all-MiniLM-L6-v2", backend="onnx")`. Does your model not have an ONNX or OpenVINO file yet? It'll be autoexported for you. Thank me later 😉
And if you don't want to export it every time, just use "push_to_hub" or "save_pretrained".
🧵
1
0
9
🔒 Another major new feature is Static Embeddings: think word embeddings like GLoVe and word2vec, but modernized. Static Embeddings are bags of embeddings that are summed together to create text embeddings, allowing for lightning-fast embeddings without any neural networks.
🧵
1
0
10
This v3.0 release has been the biggest in the Sentence Transformers history (13k lines changed, 292 files updated), and I'm very excited to see it come to fruition. I'm very much looking forward to seeing your finetuned models on
@huggingface
🧵
1
1
9
We also support Prompt Templates now! Useful for those models that always need prompts before the text (e.g. "query: ..." or "Represent this sentence for searching relevant passages: "). Learn more about it here:
1
0
9
2️⃣ Random initialization. This requires finetuning, but finetuning is extremely quick (e.g. 3 million pairs in 7 minutes). My final model was 6.6% worse than bge-base-en-v1.5, but 500x faster on CPU (yes: 500x. I'm talking 14000 sentences per second compared to 24). 🤯
🧵
1
1
9
1️⃣ via Model2Vec, a new technique for distilling a Sentence Transformer models into static embeddings. Either via a pre-distilled model with `from_model2vec` or with `from_distillation` where you do the distillation yourself. It'll only take 5 seconds on GPU & 2 minutes on CPU
🧵
1
0
9
@orionweller
@srchvrs
@n0riskn0r3ward
@spacemanidol
@memray0
@Quantum_Stat
@bo_wangbo
Sentence Transformers v3 will completely overhaul the training loop. I'm sure it'll include just about anything you'd need, from Multi-GPU w. DDP, GradCache losses, W&B/Tensorboard integration, extensive model card generation, FA2 on various models (more coming soon), etc.
1
0
9
@bo_wangbo
@JinaAI_
Looking forward to those papers! Especially your plans for v3 & collecting parallel data. On the topic of the latter, I just reformatted 10 Parallel Sentence datasets for easy use with Sentence Transformers v3:
1
0
8
5️⃣ Dataset Release
To help you out with finetuning models, I've released 50+ ready-to-go datasets that can be used with training or finetuning embedding models. Check them out here:
🧵
1
0
8
Or skip straight to the Notebook that you can run in Colab:
0
0
8
I love this; roast your (or someone else's... 👀)
@huggingface
profile with this Space by
@enzostvs
:
2
1
8
...
- Improved callback support + an excellent Weights & Biases integration
- Gradient checkpointing, gradient accumulation
- Improved model card generation
- Resuming from a training checkpoint without performance loss
- Hyperparameter Optimization
and much more!
🧵
1
1
7
🔍 Time for some sneak previews! Soon, models trained/finetuned with Sentence Transformers will automatically include detailed model cards!
In this 🧵I'll show what's included:
- Model Details, e.g. base model, sequence length, output dimensionality, training datasets, language.
1
1
8
@huggingface
And stay tuned for future updates, I've got big plans and plenty of motivation to make 'em happen. Check out the repository to keep up to date or to submit issues/feature requests/pull requests:
0
0
8
Additionally, we now support the popular INSTRUCTOR models, such as . Check out the documentation on how to use these models:
1
0
8
@bclavie
@antoinelouis_
Looking to integrate this into Sentence Transformers in the long run, but it likely has lower priority than improving CrossEncoder training & 1-st party ColBERT.
1
0
8
SetFit v1.0.2 is out to fix some v1.0 release bugs: incorrect model cards when using custom metrics, multi-output mixed with predict_proba, the "unique" sampler, and predicting polarities of gold aspect spans in SetFit ABSA models.
Check the repo here:
1
0
7
Extremely excited to have contributed this implementation! I really think Attention Sinks will be huge to bring constant memory & constant fluency to LLMs.
1
0
7
@llama_index
@LangChainAI
@deepset_ai
Check out the release blogpost by
@mixedbreadai
to learn more details:
🧵5/5
0
2
7
Would you look at that, in the meantime
@urchadeDS
uploaded a third English model:
You know models are fresh when they're still being created during the announcements😄
0
0
7
📉 A new "Cached" variant of the powerful Multiple Negatives Ranking Loss allows normal hardware to get performance that used to only be viable on multi-gpu clusters.
🐎 Community Detection is now much faster (7x speedup at 500k sentences 🤯)
1
0
7
What should I prioritize next? Cross-Encoder (i.e. reranker) training? Cross-Encoder with ONNX/OpenVINO backends? LoRA training?
5
0
7
2️⃣ OpenVINO Backend: This backend uses Intel their OpenVINO instead, outperforming ONNX in some situations on CPU.
For CPU, the ONNX int8 quantized model can reach upwards of 3x speedup compared to the default fp32! 🏎️
🧵
1
0
7
@iamrobotbear
@ClementDelangue
They do, the new models (v2.1) are all Apache 2.0, which allows for commercial use:
-
-
-
-
0
0
6